Sebastian Lobentanzer

Despite being only used for typing in the scanpy_tl module, it throws an error in the testing process. Due to mocking? Moved out of `TYPE_CHECKING` for now.

remove all manual processing and process using formatter.py

There is some room for improvement in the `schema_info.yaml` to make the LLM queries more robust. See sub-issues.

The default process does not consider annotations on the entities in the schema information. To allow behaviours more informed by the context of the entities in the graph, the schema...

The safety KG schema is currently not the enriched, but the basic one. Need to update and make sure future schemata are also enriched.

For query generation, it may be useful to have undirected relationships at the query stage; i.e., a bidirectional relationship is represented by just one directed relationship in Cypher, but the...

This issue is associated with a parent issue, which is its editor.

[Bug] LiteLLMConversation unused exception

2

**Describe the bug** The `ValueError` in `get_litellm_object()` will never be raised, because the parameter `api_key` is not optional. **Expected behavior** Either make it optional and raise the error, or remove...

# Model name Add GPT-5 to benchmark

# Model name Nemotron nano 9b v2 seems interesting # Source https://build.nvidia.com/nvidia/nvidia-nemotron-nano-9b-v2